Where to Go Next

Congratulations on making it through the fundamentals. You now have a working mental model for how AI systems learn from data, the math underneath them, and how the two biggest application areas — vision and language — actually work. That's a solid foundation for whatever direction you want to take next: hands-on coding, deeper math, a specific application area, or just being a more informed user of AI tools.

Glossary

Activation function

A non-linear function applied to a neuron's weighted sum, letting networks learn curved, complex patterns.

Backpropagation

The algorithm that efficiently computes the gradient for every parameter in a multi-layer network.

Bias (parameter)

The extra learnable constant added to a weighted sum, letting a model shift its output up or down.

Classification

A supervised learning task that predicts a category rather than a number.

Convolution

Sliding a small kernel across an image (or grid) and computing a weighted sum at each position.

Deep learning

Machine learning using neural networks with many stacked hidden layers.

Dot product

Multiplying two vectors' matching entries and summing the results; measures directional similarity.

Embedding

A learned vector representation of a token (or other item) that places similar items near each other.

Epoch

One full pass of a training algorithm through the entire training dataset.

Feature

A measurable input property used by a model to make a prediction.

Gradient descent

The algorithm that repeatedly nudges a model's parameters downhill on the loss surface to reduce error.

Hallucination

A confident but false or unfounded statement generated by a language model.

k-Nearest Neighbors (k-NN)

A classifier that predicts a label by majority vote among the k closest training examples.

Label

The correct answer associated with a training example, used in supervised learning.

Learning rate

How large a step gradient descent takes at each iteration.

Loss function

A number that measures how wrong a model's current predictions are.

Neural network

A model made of layers of connected artificial neurons.

Overfitting

When a model memorizes training data instead of learning a pattern that generalizes to new data.

Regression

A supervised learning task that predicts a continuous number.

Reinforcement learning

Learning through trial and error, guided by rewards and penalties from an environment.

Token

A basic unit of text (often a word or subword) that a language model processes.

Vector

An ordered list of numbers representing one thing so a computer can process it.

Where to go deeper (free, authoritative resources)

• Elements of AI

  A free, widely-used introductory course from the University of Helsinki and MinnaLearn, covering AI concepts with no coding required.

• Google's Machine Learning Crash Course

  Google's free course with interactive exercises, covering the ML fundamentals in more depth, including TensorFlow examples.

• Harvard CS50's Introduction to Artificial Intelligence with Python

  A free, project-based university course covering search, knowledge, uncertainty, machine learning, and neural networks with hands-on Python projects.

• 3Blue1Brown's Neural Networks series

  Outstanding visual, intuition-first video explanations of neural networks, gradient descent, and backpropagation.

• fast.ai

  A free, practical, code-first deep learning course for people who want to build real models quickly and learn the theory along the way.

If you want to keep building, this site's Courses section has practical, project-based programming courses — a great next step once you're ready to start writing code that applies what you've learned here.